Pattern Avoiding Partitions, Sequence A054391 and the Kernel Method ;

advertisement
Available at
http://pvamu.edu/aam
Appl. Appl. Math.
ISSN: 1932-9466
Applications and Applied
Mathematics:
An International Journal
(AAM)
Vol. 6, Issue 2 (December 2011), pp. 397 – 411
Pattern Avoiding Partitions, Sequence A054391 and the Kernel Method
Toufik Mansour and Mark Shattuck
Department of Mathematics
University of Haifa
31905 Haifa, Israel
toufik@math.haifa.ac.il; maarkons@excite.com
Received:March 06, 2011; Accepted: August 08, 2011
Abstract
Sequence A054391 in OEIS, which we will denote by an , counts a certain two-pattern avoidance
class of the permutations of size n . In this paper, we provide additional combinatorial
interpretations for these numbers in terms of finite set partitions. In particular, we identify six
classes of the partitions of size n , all of which have cardinality an and each avoiding two
classical patterns. We use both algebraic and combinatorial methods to establish our results. In
one apparently more difficult case, to show the result, we make use of the kernel method in
solving a system of three functional equations which arises after a certain parameter is
introduced. We also define an algorithmic bijection between the avoidance class in this case and
another which systematically replaces the occurrences of a given pattern with those of another
having the same length.
Keywords:
Pattern avoidance, set partition, kernel method
MSC 2010 No.: 05A15, 05A18
1. Introduction
If j  1 , then let s j denote the sequence (see Barcucci et al. (2000)) which counts the
permutations of size n avoiding the patterns 321 and ( j  2)1( j  3)2  ( j  1) . Letting j vary
produces different sequences, the j = 1 and j = 2 cases, for example, corresponding to the
Motzkin numbers and to enumerators of Motzkin left factors (which was shown in Barcucci et al.
(2000)). Letting j go to infinity produces the Catalan sequence, and so there is a ``discrete
397
398
Toufik Mansour & Mark Shattuck
continuity'' between the Motzkin and Catalan sequences, as noted in Barcucci et al. (2000). In the
present correspondence, we are concerned with the case j = 3 , the terms of which we will
denote by an . In particular, we will show that an also counts certain avoidance classes of set
partitions. The sequence an occurs as A054391 in Sloane and has generating function given by
a x
n 0
n
n
= 1
2x2
2 x 2  3x  1  1  2 x  3x 2
.
(1)
The first few an values, starting with n = 0 , are given by 1, 1, 2, 5, 14, 41, 123, 374,  .
If n  1 , then a partition of [n] = {1,2,, n} is any collection of non-empty, pairwise disjoint
subsets, called blocks, whose union is [n] . (If n = 0 , then there is a single empty partition which
has no blocks.) Let Pn denote the set of all partitions of [n] . A partition  is said to be in
standard form if it is written as  = B1/B2 /  , where the blocks are arranged in ascending order
according to the size of the smallest elements. One may also represent  = B1/B2 /   Pn ,
expressed in the standard form, equivalently as  =  1 2  n , wherein j  B ,1  j  n , called
j
the canonical sequential form; and, in such case, we will write  =  . For example, the partition
 = 1,5/2,3,7/4/6,8 has the canonical sequential form  = 12231424 . Note that  =  1 2  n
possesses the restricted growth property (see, e.g., Stanton and White (1986) or Wagner (1996)
for details), meaning that it satisfies the following three conditions: (i)  1 = 1 , (ii)  is onto [k ]
for some k  1 , and (iii)  i 1  max{ 1 ,  2 , ,  i }  1 for all i , 1  i  n  1 . In what follows, we
will represent set partitions as words using their canonical sequential forms and consider some
particular cases of the general problem of counting the members of a partition class having
various restrictions imposed on the order of the letters.
A classical pattern  is a member of []m which contains all of the letters in [] . We say that a
word   [k]n contains the classical pattern  if  contains a subsequence isomorphic to  .
Otherwise, we say that  avoids  . For example, a word  =  1 2  n avoids the pattern 231
if it has no subsequence  i j k with i < j < k and  k <  i <  j and avoids the pattern 1221 if
it has no subsequence  i j k   with  i =   <  j =  k . The pattern avoidance question has
been the topic of much research in enumerative combinatorics, starting with Knuth (1974) and
Simion and Schmidt (1985) on permutations and considered, more recently, on further structures
such as words and compositions. The avoidance problem can be extended to set partitions upon
considering the question on the associated canonical sequential forms. We refer the reader to the
papers by Klazar (1996), Sagan (2010), and Jelínek and Mansour (2008) and to the references
therein. We will use the following notation. If {w1 , w2 ,} is a set of classical patterns, then let
Pn ( w1 , w2 ,) be the subset of Pn which avoids all of the patterns, whose cardinality we will
denote by pn ( w1 , w2 ,) .
AAM: Intern. J., Vol. 6, Issue 2 (December 2011)
399
In this paper, we identify six classes of the partitions of [n] each avoiding a classical pattern of
length four and another of length five and each enumerated by the number an . In addition to
providing new combinatorial interpretations for a sequence, this addresses specific cases of a
general question raised by Goyt (2008), for example, regarding the enumeration of avoidance
classes of partitions corresponding to two or more patterns. Analogous results concerning the
avoidance of two patterns by a permutation have been given, for example, by West (1995). Our
main result is the following theorem which we prove in the next section as a series of
propositions.
Theorem 1.1. If n  0 , then pn (u , v) = an for the following sets (u, v) :
(1) (1212,12221)
(4) (1221,12311)
(2) (1212,11222)
(5) (1221,12112)
(3) (1212,11122)
(6) (1221,12122) .
To show this, we give algebraic proofs for cases (1), (3), (4), and (6) and find one-to-one
correspondences between cases (1) and (2) and (5) and (6). To establish (4) and (6), we make use
of the kernel method (see Banderier et al. (2002) and Hou and Mansour (2011)), in the latter case,
to solve a system of functional equations which arises once a certain parameter has been
introduced. Our bijection between cases (5) and (6) is of an algorithmic nature and
systematically replaces occurrences of 12122 with ones of 12112 without introducing 1221 .
2. Proof of the Main Result
2.1. The cases {1212,12221} , {1212,11222} , and {1212,11122}
In this section, we consider the cases of avoiding {1212,12221} , {1212,11222} , and
{1212,11122} .
Proposition 2.1. The generating function for pn (1212,12221) and pn (1212,11122) , where
n  0 , is given by
1
2x2
2 x 2  3x  1  1  2 x  3x 2
.
Proof:
Note first that each non-empty partition   Pn (1212,12221) may be decomposed as either
 = 1  , where   contains no 1's, or as  = 111 2 1 r , for some r  2 , where the  i contain
no 1 's. The  i must avoid {1212,111} if 1  i  r  1 , with  r avoiding {1212,12221} .
400
Toufik Mansour & Mark Shattuck
Furthermore, note that each letter in  i is greater than each letter in  j if i > j in order to avoid
an occurrence of 1212 . Let f ( x ) = n0 pn (1212,1222 1) x n and g ( x ) =  n 0 pn (1212,111) x n .
From the foregoing observations, we obtain the relation
x 2 f ( x) g ( x)
,
f ( x) = 1  xf ( x) 
1  xg ( x)
or
f ( x) =
1  xg ( x)
.
1  x  xg ( x )
(2)
To compute g (x) , observe that  = Pn (1212,111) must be of the form  = 1  or  = 1 1  ,
where   and   contain no 1's and avoid {1212,111}. Furthermore, all of the letters of   are
greater than all of the letters of   in the second case. This implies
g ( x) = 1  xg ( x)  x 2 g 2 ( x),
or
g ( x) =
1  x  1  2 x  3x 2
.
2x2
(3)
Substituting (3) into (2) and simplifying yields the first case above.
To compute the generating function h(x) for pn (1212,11122) , we use the same cases as we did
for finding pn (1212,11222) . In the second case, however, consider further whether or not  2
contains a repeated letter. Note that if it does, then 1 must avoid {1212,111}. Furthermore, no
letters greater than one occurring after the third 1 can be repeated. This yields the relation
h( x) = 1  xh( x) 
x2
x  2
x 
1 


h( x)1 
  x g ( x) h( x) 
1 
,
1 x
1  x  1  2 x 
 1 2x 

or
h( x ) =
1  2 x  x 2 g ( x)
,
1  3 x  x 2  x 2 (1  x) g ( x)
from which the desired result follows from (3).
AAM: Intern. J., Vol. 6, Issue 2 (December 2011)
401
There is a direct bijection showing the equivalence of the pairs {1212,12221} and {1212,11222} .
Proposition 2.2. If n  0 , then pn (1212,12221) = pn (1212,11222) .
Proof:
Let An = Pn (1212,12221) and Bn = Pn (1212,11222) . We will define a bijection f between the
sets An and Bn in an inductive manner, the cases n = 1,2,3 clear. Suppose n  4 and   An . If
 = 12 n , then   Bn and let f ( ) =  . If   12 n , then let m  1 be the smallest letter of
 which occurs at least twice. Thus, we may write  = 12 (m  1)m1m 2  m r , where r  2
and the  i contain only letters in {m  1, m  2,} but are otherwise just as in the proof of
Proposition 2.1 above.
Given a finite word w on the alphabet of positive integers having m distinct letters, let stan( w )
denote the equivalent word on [m] having all of the same relative comparisons with regard to its
positions (often called the standardization of w ). Let  r = f ( stan ( r )) and  ro be obtained
from  r by adding m to each letter. If 1  i  r  1 , then let  'i be obtained from  i by adding
 to each letter, where  is the number of distinct letters of  r . Let f ( ) be given by
f ( ) = 12 (m  1)m ro m '1 m '2  m 'r 1.
Then f ( ) belongs to Bn since  ro avoids {1212,11222} and each  'i avoids {1212,111}. One
may verify that the mapping f is a bijection, which completes the proof.
Remark: The mapping f is seen to preserve the number of blocks; thus the members of
Pn (1212,12221) and Pn (1212,11222) having the same prescribed number of blocks are
equinumerous.
2.2. The Case {1221,12311}
Let  =  1 2  n denote a partition of [n] , represented canonically. Recall that empty sums
take the value zero, by convention. To establish this case, we divide up the set of partitions in
question according to a certain statistic, namely, the one which records the length of the maximal
increasing initial run. To do so, given k  1 , let f k (x) denote the generating function for the
number of partitions  of [n] having at least k letters and avoiding the patterns 1221 and
12311 such that  1 2  k = 12 k with  k 1  k (if there is a (k  1) -st letter). We have the
following relation involving the generating functions f k (x) .
402
Toufik Mansour & Mark Shattuck
Lemma 2.3. If k  2 , then
k 1
f k ( x) = x k  x k f 1 ( x)  x k 1 f 2 ( x)  x k 1 j f j ( x),
(4)
j =2
with initial conditions f 0 ( x) = 1 and f1 ( x ) = x  x f 1 ( x ) , where f k ( x ) = i  k f i ( x ) .
Proof:
Note that f1 ( x) = x  x f 1 ( x) , since a partition in this case may just have one letter or start 11 . If
k = 2 , then a partition  enumerated by f 2 ( x) must either be 12 or start with 121 or 122 ,
which implies
f 2 ( x) = x 2  x f 2 ( x )  x 2 f 1 ( x ).
Note that in the second case, the second letter 1 is extraneous (concerning possible occurrences
of 12311 or 1221 ) and therefore can be removed without affecting the enumeration, while in the
third case, the letters 1 and 2 at the beginning are extraneous.
If k  3 , then we consider the following cases concerning the partitions enumerated by f k (x) :
(i ) 12 k ,
(ii ) 12 kj , where 1  j  k  2,
(iii ) 12 k (k  1) ,
(iv) 12 kk .
The first case contributes x k . Note that in the second case, the word   contains no letters in [ j ] ,
for otherwise there would be an occurrence of 1221 if it contained a letter in [ j  1] or an
occurrence of 12311 if it contained the letter j . Thus, the letters ( j  1)( j  2) k  , taken
together, comprise a partition of the form enumerated by f k  j (x ) , which implies the
contribution in this case is x j 1 f k  j ( x ) . Similar reasoning in the third case yields a contribution
of x k 1 f 2 ( x ) , since   can contain no letters in [k  2] and thus (k  1)k  is a partition of the
form enumerated by f 2 ( x) (the factor of x k 1 accounts for the letters 12 (k  2) as well as the
second occurrence of k  1 ). Finally, in the fourth case, no member of [k  1] can occur in   ,
with the second k extraneous, which implies a contribution of x k f 1 ( x ) . Combining all of the
cases yields (4), which is also seen to hold in the case k = 2 as well.
Proposition 2.4. The generating function for pn (1221,12311) , n  0 , is given by
AAM: Intern. J., Vol. 6, Issue 2 (December 2011)
1
2x2
2 x 2  3x  1  1  2 x  3x 2
403
.
Proof:
Define the generating function
F ( x, y ) =  k  0 f k ( x ) y k .
First note that
f 1 ( x ) = f1 ( x )  f 2 ( x )   = F ( x,1)  1 , f1 ( x ) = x  x f 1 ( x ) = xF ( x,1) , and
f 2 ( x) = F ( x,1)  f1 ( x)  1 = (1  x) F ( x,1)  1 .
Multiplying (4) by y k and summing over k  2 yields
k 1


F ( x, y ) = 1  f1 ( x) y    x k  x k f 1 ( x)  x k 1 f 2 ( x)  x k 1 j f j ( x)  y k
k 2 
j =2

2 2
2 2
2
x y
x y
xy

= 1  xyF ( x,1) 
[ F ( x,1)  1] 
[(1  x) F ( x,1)  1]
1  xy 1  xy
1  xy
 k 1

   x k 1 j f j ( x)  y k
k  2  j =2

= 1  xyF ( x,1) 
xy 2
x2 y
f j ( x) y j
[ F ( x,1)  1] 

1  xy
1  xy j  2
= 1  xyF ( x,1) 
i
xy 2
x2 y
[ F ( x,1)  1] 
(
)
f
x
yj


i
1  xy
1  xy i  2
j =2
= 1  xyF ( x,1) 
xy 2
[ F ( x,1)  1]
1  xy


x2 y  y2
y
( F ( x,1)  f1 ( x)  1) 
( F ( x, y )  f1 ( x) y  1)  ,

1  xy 1  y
1 y

which implies


1  xy  xy 2  x 2 y 2 xy(1  xy  y 2  2 xy 2 )
x2 y2
1 
 F ( x, y ) =
F ( x,1).

1  xy
(1  xy)(1  y )
 (1  xy )(1  y ) 
(5)
404
Toufik Mansour & Mark Shattuck
This type of functional equation can be solved systematically using the kernel method (see
Banderier et al. (2002)). In this case, if we assume that y = y0 in (5), where y0 satisfies
1
x 2 y02
= 0 , i.e.,
(1  xy0 )(1  y0 )
y0 =
1  x  1  2 x  3x 2
,
2 x(1  x)
then
(1  y0 )(1  xy0  xy02  x 2 y02 )
pn (12311,1221) x = F ( x,1) = 

xy0 (1  xy0  y02  2 xy02 )
n0
n
= 1
2x2
2 x 2  3x  1  1  2 x  3x 2
,
as required. (Note that F (0,1) = 1 dictates our choice of root for y0 .)
Remark: Substituting the expression above for F (x,1) into (5) recovers the expression for
F ( x, y) , from which one can compute an explicit formula for the coefficient of x n y k .
2.3. The Cases {1221,12112} and {1221,12122}
For these cases, we show first that pn (1221,12122) = pn (1221,12112) through a direct bijection
and then show that the generating function for pn (1221,12122) is given by (1) above. Before
defining the bijection, we will need the following two lemmas.
Lemma 2.5. Suppose   Pn (1221,12122) has at least one occurrence of the pattern 12112 . Let
b  2 be the smallest letter such that there exists a  [b  1] for which there is a subsequence in
 given by abaab . Then (i) the element b occurs exactly twice in  and (ii) the element a is
uniquely determined.
Proof:
For (i), suppose, to the contrary, that b occurs at least three times in  . Let  denote an
occurrence of the subsequence abaab in  ; we may assume that a and b in  correspond to
initial occurrences of letters of their kind. If an additional b occurs to the right of the third a in
 , then there would be an occurrence of 12122 in  , which is not allowed. If an additional b
occurs to the left of the third a in  , then there would be an occurrence of 1221 , which is also
not allowed. Thus, there are exactly two b 's in  and they correspond to an occurrence of
12112 . Note that the minimality of b is not needed for this part.
AAM: Intern. J., Vol. 6, Issue 2 (December 2011)
405
To show (ii), suppose, to the contrary, that there exists a subsequence  in  involving the two
b 's given by  = cbccb for some c  [b  1], c  a . Suppose first c < a . If the third c of 
occurs to the right of the second a in  above, then there would be an occurrence of 1221 , and
if it occurs to the left of the second a in  , then there would be an occurrence of 12122 . Thus,
there are no occurrences of cbccb with c < b (in fact, this shows that any letter c < a in  can
only occur prior to the left-most b in  ).
Now suppose a < c < b and let  again denote a possible occurrence of cbccb . If the third c of
 occurs after the third a in  , then there would be an occurrence of 12112 of the form acaac ,
with c < b , contradicting the assumed minimality of b . On the other hand, if the third c of 
occurs before the third a of  , then there would be an occurrence of 1221 of the form acca .
Thus, no element of [b  1] can form an occurrence of 12112 with b , which completes the proof
of (ii).
Lemma 2.6. Suppose  =  1 2  n  Pn (1221,12122) contains at least one occurrence of the
pattern 12112 and let a and b be as defined in Lemma 2.5 above. Write  = W1bW2bW3 , where
W3 is possibly empty. Then we have the following: (i) only letters greater than b can occur in
W2 , with the exception of a , and no letter other than a can occur more than once in W2 ; and (ii)
any letter occurring in W2 can occur at most once in W3 , all of whose letters are greater than b .
Proof:
To prove (i), write W2 = 1a 2 a  r 1a r for some r  3 , where the  i are possibly empty and
contain no a 's. Suppose, to the contrary, that c < b occurs in W2 , where c  a . If c < a is in  i
for some i  2 , then there would be an occurrence of 1221 , whereas if it is in 1 , then there
would be an occurrence of 12122 . If a < c < b and c belongs to  i for some 1  i < r , then
there would be an occurrence of 1221 with acca , whereas if c belongs to  r , then there would
be an occurrence of 12112 of the form acaac with c < b , contradicting the minimality of b .
Thus, only letters greater than b can occur in W2 , with the exception of a , and each letter
greater than b can occur in W2 at most once so as to avoid 1221 . Statement (ii) is also a
consequence of  belonging to Pn (1221,12122) .
We now establish the equivalence of avoiding {1221,12122} and {1221,12112} .
Proposition 2.7. If n  0 , then pn (1221,12122) = pn (1221,12112) .
406
Toufik Mansour & Mark Shattuck
Proof:
Let A = Pn (1221,12122) and B = Pn (1221,12112) . We will describe a bijection f between the
sets A and B , algorithmically, as follows. Suppose   A . If   B , then let f ( ) =  .
Otherwise,  0 =  contains at least one occurrence of the pattern 12112 . Let b0 denote the
smallest letter b in  0 for which there exists a subsequence abaab for some a < b and let a0
denote the corresponding letter a , which is uniquely determined, by Lemma 2.5. Suppose we
write  0 as  0 = W1b0W2b0W3 as in Lemma 2.6 above, where W2 = 1a0 2 a0  r 1a0 r . Let
 1 = W1b0W '2 b0W3 , where W '2 = 1a0 2b0  r 1b0 r in which we have changed all but the first
a0 occurring in W2 to b0 . Note that this replaces all of the occurrences of 12112 involving a0
and b0 with ones of 12122 . Using Lemma 2.6, one can verify that no occurrences of 1221 are
introduced.
If  1 has no occurrences of 12112 , then let f ( ) =  1 . Otherwise, let b1 denote the smallest
letter b in  1 for which there exists a subsequence abaab for some a < b . One can verify
b1 > b0 . By reasoning similar to that used in the proof of Lemma 2.5 above, one can also show
that a letter a < b1 for which the subsequence ab1aab1 occurs in  1 is uniquely determined,
which we'll denote by a1 . Let  2 denote the partition obtained by changing all of the letters a1 ,
except the first, coming after the leftmost b1 to b1 . No 1221 subsequences are introduced, which
follows from the minimality of b1 . Now repeat the above process, considering  2 .
Since b0 < b1 < b2 <  , the procedure described must end in a finite number, say t , of steps, with
the resulting partition  t belonging to Pn (1221,12112) . Let f ( ) =  t . Note that the largest b
for which there exists a < b such that ababb occurs in  t is b = bt 1 whenever t  1 . This
follows from the fact one can verify that no occurrences of 12122 in which the 2 corresponds to
a letter greater than bi are introduced in the transition from  i to  i 1 for all i . If t  1 , then one
can also verify that the largest a < bt 1 for which there is a subsequence of the form abt 1abt 1bt 1
in  t is a = at 1 . So to reverse the algorithm, we first consider the largest letter b (if it exists) for
which ababb occurs in  t for some a < b and then consider the largest such a corresponding to
this b . One can then change the letters accordingly to reverse the final step of the algorithm
describing f and the other steps can be similarly reversed, going from last to first.
Note that the above bijection preserves the number of blocks. Below we provide an example
when n = 15 :
 =  0 = 123224253647567   1 = 123234353647567   2 = 123234354647567 
 3 = 123234354657567   4 = 123234354657667   5 = 123234354657677 = f ( ).
AAM: Intern. J., Vol. 6, Issue 2 (December 2011)
407
We now find an explicit formula for the generating function for the number of partitions in
Pn (1221,12122) . In order to achieve this, we will consider the following three types of
generating functions:
•
For all k  1 , let Fk (x) be the generating function for the number of partitions
 =  1 2  n  Pn (1221,12122) such that  1 2  k = 12 k and  k 1  k . We define
F0 ( x) = 1 .
•
For all k  2 , let H k (x) be the generating function for the number of partitions
 =  1 2  n  Pn (1221,12122) such that  1 2  k = 12 k and  k 1 = 1 .
•
For all k  2 , let Gk (x) be the generating function for the number of partitions
 =  1 2  n  Pn (1221,12122) such that  1 2  k = 12 k ,  k 1 = 1 and  j  1 for
all j = k  2,, n .
We define the further generating functions F ( x, y ) =  k 0Fk ( x ) y k , H ( x, y ) = k  2H k ( x) y k , and
G ( x, y ) = k  2Gk ( x ) y k . Our goal will be to find F (x,1) , which is the generating function for the
sequence pn (1221,12122) , n  0 . The next three lemmas provide relations which we will need
between these generating functions.
Lemma 2.8. We have
F ( x, y ) = 1 
1
xy
F ( x,1) 
H ( x, y ).
1  xy
1  xy
Proof:
be
the
generating
function
for
the
number
of
partitions
Let
a (x )
 =  1 2  n  Pn (1221,12122) such that  1 2  k =  , where  is some word. By the
definitions, we have
F0 ( x) = 1,
F1 ( x) = x  a11 ( x) = x  x Fj ( x) = x  x( F ( x,1)  1),
j 1
k
k 1
j =1
j =1
Fk ( x) = x k  a12kj ( x) = x k  x j 1a12( k 1 j )1 ( x)  x k a1 ( x)
k 1
= x k  x j 1 H k 1 j ( x)  x k ( F ( x,1)  1),
j =1
Hence,
k  2.
408
Toufik Mansour & Mark Shattuck
F ( x, y ) =
1
yj
xy

H j ( x) 
( F ( x,1)  1),
1  xy j  2 1  xy
1  xy
which is equivalent to
F ( x, y ) = 1 
1
xy
F ( x,1) 
H ( x, y ),
1  xy
1  xy
as required.
Lemma 2.9. We have
H ( x, y ) = G ( x, y ) 
xy
( yH ( x,1)  H ( x, y )).
1 y
Proof:
Let us write an equation for the generating function H k (x) . Suppose  =  1 2  n is any
member of Pn (1221,12122) such that  1 2  k 1 = 12 k1 . Consider the following two cases:
(1)  j  1 for all j = k  2, k  3,, n or (2) there exists at least one index j > k  1 such that
 j = 1 . Clearly, the first case contributes Gk (x) . For the second case, we write  = 12 k1 1  ,
observing that since  avoids 1221 , there exists some  such that   = (k  1)  . Since the
members of Pn (1221,12122) of the form 12 k1(k  1) 1  are in one-to-one correspondence
with the members of Pn 1 (1221,12122) of the form 12 1  , we see that the second case
contributes x   k H  (x ) . Thus,
H k ( x) = Gk ( x)  xH  ( x),
k  2.
k
Multiplying this relation by y k , and summing over all k  2 , yields
H ( x, y ) = G ( x, y )  x 
j 2
y 2  y j 1
H j ( x),
1 y
which is equivalent to
H ( x, y ) = G ( x, y ) 
xy
( yH ( x,1)  H ( x, y )),
1 y
AAM: Intern. J., Vol. 6, Issue 2 (December 2011)
409
as required.
Lemma 2.10. We have


x2 y2
1 
G ( x, y )
 (1  y )(1  xy) 
x3 y 2
x2 y2
x 4 y 2 (2  x)
x3 y 2

G ( x,1) 
=
F ( x,1) 
H ( x,1).
1  xy (1  y )(1  xy)
(1  x)(1  xy)
(1  x)(1  xy)
Proof:
Let a (x ) (respectively, b (x) ) be the generating function for the number of partitions
 =  1 2  n  Pn (1221,12122) such that  1 2  k =  (respectively,  1 2  k =  and
 j  1 for all j  k  1 ). From the definitions, we have
k 1
Gk ( x) = x k 1  b12k 1 j ( x)  b12k1k ( x)  b12k1( k 1) ( x)
j =2
k 1
= x k 1  x j Gk 1 j ( x)  x k  2 F ( x,1)  b12k 1( k 1) ( x),
j =2
and for all   k  1 ,
b12k1( k 1) ( x)
k
= x  1  b12k1( k 1)j ( x) 
j =2
k
= x  1  x j G 1 j ( x) 
j =2

b
j = k 1
 1
x H
j = k 1
12k 1( k 1)j
j
 1 j
( x)  b12k1( k 1)(  1) ( x)
( x)  x  1 ( F ( x,1)  1)  b12k1( k 1)(  1) ( x).
Hence, by summing over all   k  1 , we obtain
x k 1
xk 2
x k 1
k 2
 x F ( x,1) 
( F ( x,1)  1) 
Gk ( x) =
H ( x,1)
1 x
1 x
1 x
k 1
k
x j  x k 1
x 2  x k 1
Gk  2 j ( x) 
 x j Gk 1 j ( x)  
G ( x).
1 x
1  x k
j =2
j =3
Multiplying the above recurrence by y k , and summing over all k  2 , yields
410
Toufik Mansour & Mark Shattuck
x3 y 2
x4 y 2
x4 y 2
( F ( x,1)  1)

G ( x, y ) =
F ( x,1) 
(1  x)(1  xy ) 1  xy
(1  x)(1  xy )

x3 y 2
x2 y
x3 y
H ( x,1) 
G ( x, y ) 
G ( x, y )
(1  x)(1  xy )
1  xy
1  xy

x2 y
x3 y 2
G ( x,1),
( yG ( x,1)  G ( x, y )) 
(1  x)(1  y )
(1  x)(1  xy )
which leads to the required result.
Now we are ready to find an explicit formula for the generating function F (x,1) for the number
of partitions in Pn (1221,12122) . Lemmas 2.8, 2.9, and 2.10 give rise to a system of functional
equations which we will solve using the kernel method (see Hou and Mansour (2011)). Lemma
2.8 implies
F ( x,1) =
1 x
1

H ( x,1),
1 2x 1 2x
and replacing y by
(6)
1
in Lemma 2.9 gives
1 x
1 

H ( x,1) = (1  x)G  x,
.
 1 x 
(7)
1  x  1  2 x  3x 2
1
Replacing y first by y1 =
and then by y 2 =
in Lemma 2.10 gives
2 x(1  x)
1 x
G ( x,1) =
x3 y12
x 4 y12 (2  x)
x3 y12

F ( x,1) 
H ( x,1),
1  xy1 (1  x)(1  xy1 )
(1  x)(1  xy1 )
1  x 3 y2 x 4 y22 (2  x)
x 3 y22
x

=
(
,1)
G  x,
F
x
H ( x,1) 
G ( x,1).



1  3x
1  3x
1  3x
 1  x  1  3x
Solving the system of equations (6)-(9) yields the following result.
Proposition 2.11. The generating function for pn (1221,12122) , n  0 , is given by
F ( x,1) = 1 
2x2
2 x 2  3x  1  1  2 x  3x 2
.
(8)
(9)
AAM: Intern. J., Vol. 6, Issue 2 (December 2011)
411
3. Conclusion
In this paper, we have identified six subsets of the partitions of size n, each avoiding a classical
pattern of length four and another of length five and each enumerated by the sequence an. We
have thereby obtained new classes of combinatorial structures enumerated by an, apparently the
first such examples related to set partitions. Furthermore, numerical evidence shows that there
are no other members of the (4, 5) Wilf-equivalence class for set partitions corresponding to the
sequence an. We have used both algebraic and combinatorial methods to establish our results. In
a couple of the seemingly more difficult cases, we make use of the kernel method to solve the
functional equations that are satisfied by the related generating functions. It would be interesting
to see if any of the other sequences in the ''discrete continuity'' mentioned in the introduction
enumerate restricted subsets of partitions when j > 3 (such as those avoiding three or more
patterns). Finally, it seems that the technique of introducing an auxiliary parameter and solving
the functional equations which arise as a result using the kernel method would have wider
applicability to other questions of avoidance not only for set partitions but also for other finite
discrete structures, such as k-array words or permutations.
REFERENCES
Banderier, C., Bousquet-Mélou, M., Denise, A., Flajolet, P., Gardy, D. and GouyouBeauchamps, D. (2002). Generating functions for generating trees, (Formal Power Series
and Algebraic Combinatorics, Barcelona, 1999), Discrete Math. 246:1-3, 29-55.
Barcucci, E., Lungo, A.D., Pergola, E. and Pinzani, R. (2000). From Motzkin to Catalan
permutations, Discrete Math. 217, 33-49.
Goyt, A. (2008). Avoidance of partitions of a three-element set, Adv. in Appl. Math. 41, 95-114.
Hou, Q.H. and Mansour, T. (2011). Kernel method and systems of functional equations with
several conditions, J. Comput. Appl. Math. 235:5, 1205-1212.
Jelínek, V. and Mansour, T. (2008). On pattern-avoiding partitions, Electron. J. Combin. 15, #
R39.
Klazar, M. (1996). On abab -free and abba -free set partitions, European J. Combin. 17, 53-68.
Knuth, D.E. (1974). The Art of Computer Programming, Vol's. 1 and 3, Addison-Wesley,
Reading, Massachusetts.
Sagan, B.E. (2010). Pattern avoidance in set partitions, Ars Combin. 94, 79-96.
Simion, R. and Schmidt, F.W. (1985). Restricted permutations, European J. Combin. 6, 383-406.
Sloane, N.J. The On-Line Encyclopedia of Integer Sequences, http://oeis.org.
Stanton, D. and White, D. (1986). Constructive Combinatorics, Springer, New York.
Wagner, C. (1996). Generalized Stirling and Lah numbers, Discrete Math. 160, 199-218.
West, J. (1995). Generating trees and the Catalan and Schröder numbers, Discrete Math. 146,
247-262.
Download